[Gluster-users] geo-replication loops

[Giovanni Toraldo]

Hi,

I was thinking about a common (I hope!) use case of Glusterfs geo-replication.

Imagine 3 different facility having their own glusterfs deployment:
* central-office
* remote-office1
* remote-office2

Every client mount their local glusterfs deployment and write files
(i.e.: user A deposit a PDF document on remote-office2), and it get
replicated to the central-office glusterfs volume as soon as possibile
using geo-replication feature. After a while, that PDF document should
be replicated to remote-office1 using geo-replication feature.
Hopefully an user B on remote-office1 should be able to upload new
files that get replicated to central-office and remote-office2 too.

In short, this is the scenario I would like to achieve with GlusterFS
geo-replication:
remote-office1 <--> central-office <--> remote-office2

Obviously, it's possible to create an unique volume among all
facilities, but they are connected with 2mbits HDSL and we don't want
that user applications get stuck if the network is saturated or
unreliable.

Does this scenario is feasible?

Every suggestion is highly appreciated, thanks!

-- 
Giovanni Toraldo - LiberSoft
http://www.libersoft.it